[
https://issues.apache.org/jira/browse/PHOENIX-4921?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16629824#comment-16629824
]
Lars Hofhansl commented on PHOENIX-4921:
----------------------------------------
Hmm... I'm not sure how much I like this early peek() in the first place (even
in the parallel case).
It seems slightly dubious to start the scanner already at the time of enqueuing
with the ThreadPool instead of when a thread actually picks the respective task
up. I'm curious what the actual gain is.
With the defaults the queue size is 5000 with 128 threads, it might be a while
before a thread actually gets to pick up the scanner and use. All the while the
scanner has an open lease in HBase and might even time out. I think
PHOENIX-4927 might help with that part, though.
> Change SerialIterators to call peek() so that we can detect if the table
> split and we need to reload the table region cache.
> ----------------------------------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-4921
> URL: https://issues.apache.org/jira/browse/PHOENIX-4921
> Project: Phoenix
> Issue Type: Bug
> Reporter: Thomas D'Silva
> Assignee: Thomas D'Silva
> Priority: Major
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
